Impaired interhemispheric synchrony in patients with iridocyclitis and classification using machine learning: an fMRI study.

BACKGROUND

This study examined the interhemispheric integration function pattern in patients with iridocyclitis utilizing the voxel-mirrored homotopic connectivity (VMHC) technique. Additionally, we investigated the ability of VMHC results to distinguish patients with iridocyclitis from healthy controls (HCs), which may contribute to the development of objective biomarkers for early diagnosis and intervention in clinical set.

METHODS

Twenty-six patients with iridocyclitis and twenty-six matched HCs, in terms of sex, age, and education level, underwent resting-state functional magnetic resonance imaging (fMRI) examinations. The study employed the voxel-mirrored homotopic connectivity (VMHC) technique to evaluate interhemispheric integration functional connectivity indices at a voxel-wise level. The diagnostic efficacy of VMHC was evaluated using a support vector machine (SVM) classifier, with classifier performance assessed through permutation test analysis. Furthermore, correlation analyses was conducted to investigate the associations between mean VMHC values in various brain regions and clinical features.

RESULTS

Patients with iridocyclitis exhibited significantly reduced VMHC signal values in the bilateral inferior temporal gyrus, calcarine, middle temporal gyrus, and precuneus compared to HCs (voxel-level P < 0.01, Gaussian Random Field correction; cluster-level P < 0.05). Furthermore, the extracted resting-state zVMHC features effectively classified patients with iridocyclitis and HCs, achieving an area under the receiver operating characteristic curve (AUC) of 0.74 and an overall accuracy of 0.673 (P < 0.001, non-parametric permutation test).

CONCLUSION

Our findings reveal disrupted interhemispheric functional organization in patients with iridocyclitis, offering insight into the pathophysiological mechanisms associated with vision loss and cognitive dysfunction in this patient population. This study also highlights the potential of machine learning in ophthalmology and the importance of establishing objective biomarkers to address diagnostic heterogeneity.